Houdini with 1:1 Robbolito-code?

[kranium]

Well, Robbolito is pubic domain, right? So there would be nothing wrong with this.

The Robbolito versions that are being used in the code comparison above are 0.085g3 and 0.09.
These have been meticulously translated to common English programming terms.
The are greatly modified, improved, stronger, and bug-free versions (with lots of original coding and ideas) of the very original ippolit.wikispaces.com Robbolito releases.

They were released by myself and Sentinel under GPL license from www.chesslogik.com.

[hgm]

And is this part essentially different from the public-domain version?

[kranium]

And is this part essentially different from the public-domain version?

If you compare one of these with the original wikispaces release, they are like night and day...significantly stronger, completely bug free, with code that is much easier for any chess programmer to follow.

[hgm]

Well, I don't want to compare them as chess programs. So it is totally irrelevant whether they are stronger, bug free, better layed out, etc.

The question was if the particular piece of code that was shown and compared to Houdini was the same. I.e. did any of the shown parameters undergo chages related to the strength increase, and where there originally bugs in this part of the code. I suppose that since Houdini is not available in source, the comparison must have been done throughdecompiling it, so layout would not play a role at all..

[SchachProfi]

Thanks Carlos, guess for most this is indeed a much more readable version... on the other side, those understanding x86-assembly can also deal with plain-text...

Anyways, its no real surprise to me, its too obvious...
- RH, an unknown developer, climbed the top of the engine-league out of nowhere... where other take years and years to improve their versions.
- RH is a regular poster on the immortal "home forum" of the Ippolit-development
- The UCI output is nearly identical to Robbolito
- Simply change the queen-eval, multiply the calculation-evals you see with 2 and see how pretty similar H.'s evals get with Robbolito/IvanHoe...
- Not to speak about similarities in the disassembled code itself...
- etc. etc. etc.

[Mike S.]

You make me sick.

[SchachProfi]

And is this part essentially different from the public-domain version?

If you compare one of these with the original wikispaces release, they are like night and day...significantly stronger, completely bug free, with code that is much easier for any chess programmer to follow.

Honestly i would not even be surprised if he has chosen your Robbolito as base, i myself was impressed by its strength when i saw it ~2 years back... and when it comes in English-conventions/namings, its for sure a perfect start...

[kranium]

Well, I don't want to compare them as chess programs. So it is totally irrelevant whether they are stronger, bug free, better layed out, etc.

The question was if the particular piece of code that was shown and compared to Houdini was the same. I.e. did any of the shown parameters undergo chages related to the strength increase, and where there originally bugs in this part of the code. I suppose that since Houdini is not available in source, the comparison must have been done throughdecompiling it, so layout would not play a role at all..

Well the 1st thing I can think of regarding 'bugs' is that the original Robbolito's UCI output used long long integers and was causing issues.

Basically I fixed it by type defining an unsigned 64-bit integer, and called it I64u...

RobboLito 0.085f1:
"info time %lld nodes %lld nps %lld cpuload %d\n"

RobboLito 0.085g3
"info time %I64u nodes %I64u nps %I64u cpuload %u"

When examining Houdini_w32_1CPU.exe (build 2010-05-16), you can find:
"info time %I64u nodes %I64u nps %I64u cpuload %u"

I posted previously about this and asked Robert if would explain:
http://talkchess.com/forum/viewtopic.ph ... =&start=70
but never got an answer

[SchachProfi]

Norm's link for all reading as guest: http://talkchess.com/forum/viewtopic.ph ... 93&t=40685

I would also be interested into RH's answers to this one:
http://talkchess.com/forum/viewtopic.ph ... 53&t=40685

Well, but sadly we won't get one, or just the usual word-dances...

[mar]

Well the 1st thing I can think of regarding 'bugs' is that the original Robbolito's UCI output used long long integers and was causing issues.

Basically I fixed it by type defining an unsigned 64-bit integer, and called it I64u...

I thought you meant bugs